计算机科学

计算机科学 ›› 2026, Vol. 53 ›› Issue (10): 295-300.doi: 10.11896/jsjkx.250900130

• 人工智能 • 上一篇    下一篇

CLERAG:基于认知层次评估的法律问答检索增强生成方法

赵旭楠, 赵卓峰   

  1. 北方工业大学人工智能与计算机学院 北京 100144
  • 收稿日期:2025-09-21 修回日期:2025-12-25 发布日期:2026-10-13
  • 通讯作者: 赵卓峰(edzhao@ncut.edu.cn)
  • 作者简介:(2806717850@qq.com)
  • 基金资助:
    国家重点研发计划(2023YFC3107900);广州市重点研发计划(202206030009)

CLERAG:Retrieval-augmented Generation Framework for Legal QA with Cognitive-levelEvaluation

ZHAO Xunan, ZHAO Zhuofeng   

  1. School of Artificial Intelligence and Computer Science,North China University of Technology,Beijing 100144, China
  • Received:2025-09-21 Revised:2025-12-25 Online:2026-10-13
  • About author:ZHAO Xunan,born in 2000,postgra-duate,is a member of CCF(No.A02977G).Her main research interest is large-scale model engineering applications.
    ZHAO Zhuofeng,born in 1977,Ph.D,researcher,Ph.D supervisor,is a member of CCF(No.08923D).His main research interests include large-scale model engineering applications and embodied intelligence.
  • Supported by:
    National Key Research and Development Program of China(2023YFC3107900) and Guangzhou Key Research and Development Program(202206030009).

PDF (PC)

58

摘要： 随着大语言模型在法律问答领域的广泛应用,现有方法在语义理解、法条匹配与推理链条构建方面暴露出诸多问题。为此,设计了一种基于认知层次评估的检索增强生成方法CLERAG,引入认知外包思想,将语义理解与法条筛选流程化拆分,通过外部模型协作完成检索增强生成。CLERAG包括面向认知的查询增强与“发散-收敛”式的认知层次评估两部分。前者通过基于法律认知结构和约束的标准化重写与关键词扩展,将用户查询转换为规范化法律表达,提高检索相关性与覆盖率;后者模拟法律专家“发散-收敛”认知路径,结合语义评分和验证模型进行法条筛选,并通过融合策略确定最适用的法条片段。在CivGuideQA与LawBench两个中文法律问答数据集上进行了实验,结果显示,CLERAG在多个指标上均超越现有基线,在CivGuideQA数据集上,其ROUGE-L指标相较于标准RAG方法提升了13.30%。

关键词: 法律问答, 检索增强生成, 查询增强, 大语言模型, 认知层次评估

Abstract: With the growing use of large language models in legal question answering,existing methods still struggle with semantic understanding,provision matching,and reasoning.To address these issues,this paper proposes CLERAG,a retrieval-augmented generation method based on cognitive-level evaluation.Inspired by cognitive outsourcing,CLERAG structures semantic understanding and provision filtering as cooperative processes with external models.It consists of two components:cognitive-oriented query enhancement,which employs standardized rewriting and keyword expansion to align user queries with legal expressions,and a “divergent-convergent” evaluation that combines semantic scoring and a verification model to identify the most applicable provisions.Experimental results on CivGuideQA and LawBench show that CLERAG consistently outperforms baseline methods,achieving a 13.30% improvement in ROUGE-L over standard RAG on CivGuideQA.

Key words: Legal question answering, Retrieval-augmented generation, Query enhancement, Large language model, Cognitive-level evaluation

中图分类号: 

  • TP181
[1] AN Z W,LAI Y X,FENG Y S.Natural Language Understan-ding for Legal Text:A Review[J].Journal of Chinese Information Processing,2022,36(8):1-11.
[2] ZHANG H,ZHENG W H,DOU Z C,et al.Integrating Multi-Layer Structure Information of Law for Legal Judgement Prediction[J].Computer Engineering and Applications,2023,59(3):253-263.
[3] ACHIAM J,ADLER S,AGARWAL S,et al.Gpt-4 technical report[J].arXiv:2303.08774,2023.
[4] BAI J,BAI S,CHU Y,et al.Qwen technical report [J].arXiv:2309.16609,2023.
[5] BROWN T,MANN B,RYDER N,et al.Language models are few-shot learners[J].Advances in Neural Information Proces-sing Systems,2020,33:1877-1901.
[6] JI Z,LEE N,FRIESKE R,et al.Survey of hallucination in natural language generation[J].ACM Computing Surveys,2023,55(12):1-38.
[7] LEWIS P,PEREZ E,PIKTUS A,et al.Retrieval-augmentedgeneration for knowledge-intensive nlp tasks[J].Advances in Neural Information Processing Systems,2020,33:9459-9474.
[8] MA X,GONG Y,HE P,et al.Query rewriting in retrieval-augmented large language models[C]//Proceedings of the 2023 Conference on Empirical Methods in Natural Language Proces-sing.2023:5303-5315.
[9] LI R,WANG Y,WEN Z,et al.Different paths to the same destination:Diversifying LLMs generation for multi-hop open-domain question answering[J].Knowledge-Based Systems,2025,309:112789.
[10] ASAI A,WU Z,WANG Y,et al.Self-rag:Learning to retrieve,generate,and critique through self-reflection[J].arXiv:2310.11511,2023.
[11] BOSTROM N,SANDBERG A.Cognitive enhancement:me-thods,ethics,regulatory challenges[J].Science and Engineering Ethics,2009,15(3):311-341.
[12] WANG F Z,TANG X Y,YU S Q.Cognitive outsourcing based on generative artificial intelligence:An Analysis of interactive behavioral patterns and cognitive structural features[J].Acta Psychologica Sinica,2025,57(6):967.
[13] FEI Z,SHEN X,ZHU D,et al.Lawbench:Benchmarking legal knowledge of large language models[C]//Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing.ACL,2024:7933-7962.
[14] WANG Y,SHEN X,HUANG Z,et al.cLegal-QA:a Chinese legal question answering with natural language generation me-thods[J].Complex & Intelligent Systems,2025,11(1):77.
[15] DENG W,PEI J,KONG K,et al.Syllogistic reasoning for legal judgment analysis[C]//Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing.2023:13997-14009.
[16] ZHOU Z,SHI J X,SONG P X,et al.Lawgpt:A chinese legal knowledge-enhanced large language model[J].arXiv:2406.04614,2024.
[17] LI H,AI Q,CHEN J,et al.SAILER:structure-aware pre-trained language model for legal case retrieval[C]//Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval.2023:1035-1044.
[18] HUANG Q,TAO M,ZHANG C,et al.Lawyer llama technical report[J].arXiv:2305.15062,2023.
[19] GUU K,LEE K,TUNG Z,et al.Retrieval augmented language model pre-training[C]//International Conference on Machine Learning.PMLR,2020:3929-3938.
[20] MAO S,JIANG Y,CHEN B,et al.RaFe:Ranking Feedback Improves Query Rewriting for RAG[C]//Findings of the Association for Computational Linguistics(EMNLP 2024).ACL,2024:884-901.
[21] WANG Y,LI P,SUN M,et al.Self-Knowledge Guided Retrieval Augmentation for Large Language Models [C]//Findings of the Association for Computational Linguistics(EMNLP 2023).ACL,2023:10303-10315.
[22] GAO L,MA X,LIN J,et al.Precise zero-shot dense retrieval without relevance labels[C]//Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics.ACL,2023:1762-1777.
[23] SHI W,MIN S Y.REPLUG:Retrieval-Augmented Black-Box Language Models[C]//Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies.ACL,2024:8371-8384.
[24] KARPUKHIN V,OGUZ B,MIN S,et al.Dense Passage Retrieval for Open-Domain Question Answering[C]//Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing(EMNLP).ACL,2020:6769-6781.
[25] XIONG L,XIONG C,LI Y,et al.Approximate nearest neighbor negative contrastive learning for dense text retrieval [J].arXiv.2007.00808,2020.
[26] KIM K,LEE J Y.RE-RAG:Improving Open-Domain QA Performance and Interpretability with Relevance Estimator in Retrieval-Augmented Generation[C]//Proceedings of the 2024 Conference on Empirical Methods in Natural Language Proces-sing.ACL,2024:22149-22161.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市两江新区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发